Are Algorithms the Hiring Managers of the Future?
In the war for talent, companies are looking for an edge as they recruit new employees. Some innovative firms are combining data and machine learning to streamline their hiring processes and replace the traditional resume screens of days past.
The war for talent has continued to grow across industries. However, the existing hiring process is flawed and outdated. Most companies still rely on traditional resume screens to identify candidates and recruit new employees; however, this evaluation does not reliably predict an individual’s future fit with a job or company. Instead, researchers suggest that successful placement depends on a candidate’s “fittingness” and would require hiring managers to take into account other off-resume factors.1 With the average job opening attracting 250 resumes, this requires a large time investment for HR teams to review and filter.2 In order to adapt to this new environment, some companies are reevaluating the way they identify and attract candidates.
MACHINE LEARNING IN RECRUITING
One way companies are disrupting the traditional recruiting process is through the use of data and predictive analytics. Pymetrics is a machine learning company that uses data to evaluate and recommend candidates based on their proprietary algorithms. This is a new model for job applicants. Instead of filling out the traditional resume and demographic profiles that are often included in standard job applications. Individuals play online neuroscience games that may remind you more of playing Nintendo than applying for a job. These games have all been designed using neuroscience research to reveal underlying capabilities of the test taker.
In one challenge, the user is asked to select the “missing piece” in geometric patterned puzzles. After approximately 5 min of “testing” a user receives a high-level indication on what skill was tested and how she fared. In this case, the game was measuring Pattern Recognition. In each game, pymetrics measures an individual’s behavior and collects valuable data on each user. After completing a set of 12 required games, test takers receive a personalized report on their capabilities. They also receive job matches based on pymetrics’ recommendation algorithm that uses technology similar to Netflix, Amazon, or Pandora.3 In some cases, a user might even find they are a good fit for one of pymetrics’ partner employers and have a jump-start on the application process.
By using data science and machine learning, pymetrics is able to capture nonlinear relationships between behaviors and predict success in future job functions and companies. Pymetrics is able to provide value to employers by creating algorithms that will specifically predict an individual’s fit and future success in role. From this platform, employers gain access to a broad talent pool that has already been assessed and vetted as a potential match. From the user perspective, individuals gain free skill assessment based on neuroscience games and access to high potential job matches. Overall this platform strategy utilizes data and predictive analytics to streamline the cumbersome recruiting process.
IMPLICATIONS FOR DIVERSITY AND INCLUSION
Companies who work with pymetrics to utilize machine learning in their hiring process can significantly reduce the time that HR teams spend screening resumes and identifying candidates. In addition, this process helps to open up opportunities to individuals who come from less traditional backgrounds. Similarly, this process can help to promote gender and ethnic diversity by removing human bias. By focusing on skill-based data, pymetrics is able to level the playing field in order to open opportunities to a broader set of potential candidates.4
POTENTIAL RISKS AND WATCH OUTS
While machine learning can help to streamline the recruiting process, there are also potential risks. Just as every human makes mistakes, algorithms are imperfect and error prone. Recently, Amazon backtracked on their own experiment with machine learning and hiring after discovering inadvertent gender bias.5
As companies, like pymetrics, design new models for hiring it is important to continue to be thoughtful about the way algorithms are written and what data is used. Algorithms – like people – have the potential for bias, which makes it ever more important to take a critical eye to the methodology and approach to the way they are designed.6
We must also consider which aspects of the hiring process may be missed by an algorithm. For instance, can data truly understand the underlying intentions of a job applicant or adequately measure potential interest? By better understanding these limitations companies can better integrate data-driven processes with human interactions in order to successfully identify and hire the best talent for their organization.
[Word Count: 726]
[1] Tse, Terence, Mark Esposito, and Olaf Groth. 2018. “Resumes Are Messing Up Hiring”. Harvard Business Review. https://hbr.org/2014/07/resumes-are-messing-up-hiring.
[2] Gladstone, Jennifer. 2017. “60 Hiring Statistics You Need To Know For 2017”. Ebiinc.Com. https://www.ebiinc.com/resources/blog/hiring-statistics.
[3] “pymetrics: Using Neuroscience and Data Science to Revolutionize Talent Management”. https://pymetrics.com.
[4] Gupta, Nidhi. 2018. “Three Ways Machine Learning Is Improving The Hiring Process”. Forbes. https://www.forbes.com/sites/forbestechcouncil/2018/03/26/three-ways-machine-learning-is-improving-the-hiring-process/#43dbcad90e8b.
[5] “Amazon Scraps A Secret A.I. Recruiting Tool That Didn’t Like Women”. 2018. CNBC. https://www.cnbc.com/2018/10/10/amazon-scraps-a-secret-ai-recruiting-tool-that-showed-bias-against-women.html.
[6] Tugend, Alina. 2018. “The Commonality Of A.I. And Diversity”. Nytimes.Com. https://www.nytimes.com/2018/11/06/business/dealbook/the-commonality-of-ai-and-diversity.html.
Even though I agree that this is a great tool to assess candidates general skill set and narrow down the number of applicants, I still think that Pymetrics has a long way to go in the candidate selection and assessment industry. First, it seems that they are providing a general aptitude test, something that lots of companies already do. Yes, maybe their algorithms are better at matching and predicting future success for candidates, but I don’t see a highly differentiating feature yet. Second, they are completely leaving past experience out of the picture. For positions where past experience is not relevant (i.e. entry level), this could be a great tool. However, for more senior positions, as a hiring manager, I would not rely on this tool to screen out candidates for me. What if the candidate had a bad day and scored low in a couple of parts of the test? I would advise Pymetrics to develop algorithms that could also take into account or screen candidates’ resumes for relevant experience.
I think the risks you lay out are extremely relevant, particularly the point around biases. For example, I can see the algorithm placing candidates into jobs purely based on one’s prior experiences and not actual talent given the time spent developing that particular skill. That being said, I think there is potential for this tool to be used as an initial screen but then used in tandem with an in-person interview to account for some of the potential bias as the funnel narrows.
Great response! I loved reading this particularly in light of the conversations we’ve had in both TOM and LEAD. I agree with the risks and limitations you mentioned, but I wonder how a program like this would assess the positive elements of culture. Bias is definitely a risk in the current evaluative process, but as we’ve seen in LEAD, there are real performance benefits to achieving some degree of certainty that your candidate “fits” with the culture of your organization. Can a machine be trained to assess that as well or will that always be left up to a biased human observer? Are there industries where this wouldn’t prove useful based on the types of job roles required? Is there a ceiling for its usefulness (e.g. could a company use it to find its next CEO)? I agree that innovations like these offer value as companies become more aware of potential biases in their own hiring processes, but I share your concern that they will never be a fool-proof substitute for human assessment – as biased as it can be.
This sounds like really interesting technology — but certainly comes with the pitfalls mentioned in the article and in the comments. I wonder if Pymetrics would think about switching (or expanding) their customer from HR professionals to those starting out their job search; it seems to me that it could be more successful as an individual diagnostic tool for how you think and what type of role you would thrive in. Hiring for a role from the HR perspective–especially a senior role–still requires multiple perspectives and in-person conversations to truly understand if a given candidate is the right “fit.” Until machines can master EQ like (some!) humans can, there will always be gaps in how an ML-based tool assesses potential candidates.
I think algorithm-enabled hiring tools are becoming increasingly popular for organizations that are looking for ways to remain competitive in attracting and retaining talent. I am skeptical of algorithmic solutions for predicting “fit.” I would want organizations to test these tools on high-performing employees and compare results with competitive candidates. This would be an interesting way of assessing whether or not the algorithm’s definition of “fit” is in line with the organization’s culture.
Thank you for this interesting piece, especially given our recent LEAD-like conversations across most of our classes around identifying leaders! I’m curious as to how Pymetrics is receiving and processing data to refine its machine learning algorithm — to your point about preventing bias influences, are companies providing Pymetrics information back as to whether their recommendations were hired and have stayed for X duration at a company? I ask because I wonder how much of the hiring process current Pymetrics clients attribute to Pymetrics. To ABP’s point, I can see Pymetrics being more useful for the junior population, who are still trying to build those gamified skills, whereas qualities like adaptability, resilience, negotiation, etc. are harder to assess through such an app.
Great article and some great comments so far! I agree that the technology can be valuable in certain applications, but do think that there are many limitations (many of which have already been mentioned in the comments). I also wonder what type of governance would be needed or effective if these types of systems were used. You could imagine that when this methodology is used for hiring, it could still be “sidestepped” by people networking into a position through people they know in the company. Should this be allowed, or does this give unfair advantage to those who “have an in” at the company?
Thank you for sharing a great article! It is always very intriguing to think about how far AI can replace traditionally “human” work, such as HR management, advertising, etc. I agree with your questioning in the last; We also really have to consider not only what AI can do but also what AI can NOT do. I suppose that with this algorithm they may be able to capture and predict “capability/skill sets” of applicants but may be unable to capture “true motivation” or “interpersonal skills”, which are also essential factors when it comes to identify the best potential talents. So I think that the companies should make it clear what they should benefit from this service and what they should still keep doing in recruiting, then they should combine both the ways to avoid missing out good talents. End of the day, “judgement” or “final decision-making” is still left to humans, I believe.
Thanks for the interesting piece, Catherine. I believe that it is a very interesting technology, but my main concern is that given the level of specialization required for most positions, even at an entry level, the company must develop a solution that takes into account parameters like education, past work experience and potentially extracurriculars. By doing that, their value proposition toward HR departments across industries, functions, and geographies will be much more compelling. The challenge that may arise, though, is how will the company go about getting recruiters to adopt this bold new model of hiring and overcome their fears of eventually being made obsolete.